Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Always-on Linear session discipline and one-time soft nudge for untracked repos. Use whenever non-trivial work is happening in a repo that already has a Linear tracking protocol in CLAUDE.md, AGENTS.md, or .cursor/rules; or when opening/working in a repo with no protocol and no .linear-tracking-skip file (ask once to run linear-setup). Enforces search-before-create, lifecycle, and close-the-loop. NOT for full bootstrap, sync, or monitor runs.
Use when brainstorming ideas, features, or directions for a project where independent perspectives from different model families (Claude/Codex/Gemini) would surface blind spots and spark creative options the user hasn't considered — especially "what cool things can I add", "what should I build next", "give me ideas for X"
Manage project learnings — small, atomic, validated patterns that the agent should remember across all skills and sessions. Different from /money-save (which captures full session state); learnings are individual insights that get auto-loaded into every other money-* skill's context. Use when the user has just discovered something worth remembering — a customer pattern, a pricing insight, a channel that works, a failure mode. Triggered by: 'remember this', 'log a learning', 'this is a pattern', 'show learnings', 'what have we learned', '记住这个', '存入经验', '查看经验库'.
Run all four review skills (investor, customer, operator, skeptic) on a business plan in sequence, then auto-decide where verdicts agree and surface only disagreements + taste decisions to the user. One command = full review gauntlet, intelligently filtered. Use when a plan is ready to be stress-tested before commit. Triggered by: 'panel review', 'review panel', 'run all reviews', 'review gauntlet', '审议会', '四方评审'.
Save the current business state to disk so future Claude Code sessions can pick up where you left off. Captures confirmed conclusions, ruled-out directions, open hypotheses, and the next recommended skill. Use when the user has just reached a conclusion in /money-discover, /money-strategy, /money-diagnose, or any pipeline phase, and wants to lock the state in. Also triggered by: 'save this', 'checkpoint', 'remember this', 'lock it in', '保存', '存档', '记下来', '这个结论留着'.
Codify a successful workflow from the current conversation into a permanent project-local skill. After you discover (e.g.) a cold-email sequence that converts, an ad-creative format that beats baseline, or a content-pipeline cadence that compounds — /money-skillify writes it as a SKILL.md under the project's local skills directory so future sessions can re-invoke it without re-discovery. Use when the user says: 'this worked, save it', 'codify this', 'turn this into a skill', '把这个固化下来', '存成 skill'.
Adds runtime safety rails to LLM applications with NVIDIA NeMo Guardrails, configured through Colang 2.0 flows. Covers jailbreak and prompt-injection detection, self-check input/output validation, retrieval-based fact-checking, hallucination detection, PII filtering via Presidio, toxicity detection via ActiveFence, and LlamaGuard integration. Use when adding programmable safety rules to a production LLM app, blocking jailbreaks or prompt injection, filtering PII from model inputs and outputs, verifying generated claims against retrieved sources, or tuning false positives and latency of guardrail checks. For standalone moderation only, use LlamaGuard or the OpenAI Moderation API instead.
User-facing NemoClaw guidance for installing, configuring, operating, securing, monitoring, and troubleshooting NemoClaw sandboxes. Use when users ask about NemoClaw quickstarts, OpenClaw and OpenShell relationships, local inference, remote GPU deployment, sandbox lifecycle, network policy, security posture, agent skills, command reference, or issue triage instructions.
Typed knowledge graph for structured agent memory and composable skills. Use when creating/querying entities (Person, Project, Task, Event, Document), linking related objects, enforcing constraints, planning multi-step actions as graph transformations, or when skills need to share state. Trigger on "remember", "what do I know about", "link X to Y", "show dependencies", entity CRUD, or cross-skill data access.
Build stateful chatbots with OpenAI Assistants API v2 - Code Interpreter, File Search (10k files), Function Calling. ⚠️ Sunset H1 2026; use openai-responses for new projects. Use when: building stateful chatbots, implementing RAG, or troubleshooting "thread has active run", vector store delays, polling timeouts, or file upload errors.
Implements Manus-style file-based planning for complex tasks. Creates task_plan.md, findings.md, and progress.md. Use when starting complex multi-step tasks, research projects, or any task requiring >5 tool calls. Now with automatic session recovery after /clear.
Author, scope, and operate Polylane automations — unattended agent runs wired as trigger → instructions → tools → actions → destinations. Use when the user wants something to happen without them — "whenever X happens", "on every deploy", "every morning", "auto-rollback if it breaks" — including reacting to alerts, incidents, deploys, pushes, cron schedules, or webhooks; rolling back bad deployments automatically; auto-triaging alerts into issues; scheduled reports and recurring production health checks; incident-response automation; and routing results to Slack, email, or webhooks. Retrieve from Polylane docs — never guess commands or schemas.
Use when setting up or starting a project and you want to install the optimal skill stack for it. Triggers on phrases like: "set up skills for this project", "install project skills", "what skills should I use", "power engineer setup", "configure my skill stack", "power engineer frontend", "power engineer status", or any request to set up Claude Code skills for a new or existing project. Also triggers on targeted requests like "power engineer backend", "power engineer devops", "power engineer ai", "power engineer data", "power engineer docs", "power engineer mobile", "power engineer quick", "power engineer catalog", "power engineer update", "power engineer configure", "power engineer drift". Even if the user just says "/power-engineer frontend" or similar shorthand, this skill handles it.
The meta-skill that powers all other AI tools. Prompt engineering for creative applications is the art and science of communicating with AI models to produce exactly what you envision—in images, video, audio, and text. This isn't just "write better prompts." It's understanding how different models interpret language, how to structure requests for different modalities, how to iterate systematically, and how to build prompt libraries that encode your creative vision. The best prompt engineers have developed intuition for what words trigger what responses in each model. This skill is foundational—it amplifies the effectiveness of every other AI creative skill. Master this, and you master the interface to all AI creation. Use when "prompt, prompting, prompt engineering, better prompts, prompt optimization, how to prompt, prompt strategy, prompt library, prompt template, make AI understand, prompt-engineering, prompting, meta-skill, ai-creative, foundational, optimization, iteration" mentioned.
Execute prompts as the Qwen seat through the Pi CLI on OpenRouter, with Qwen3.8 27B pinned per invocation (qwen/qwen3.8-2.4t-a95b, 256K-token context, the open-weight VLM of the Qwen3.8 Max family). Use when users explicitly request Qwen execution, when a workflow needs an independent Qwen seat, or when a cross-runner workflow wants a Qwen-backed perspective without leaving the current workspace.
Inspect an agent codebase and write a markdown plan wiring Langfuse tracing/observability (session/trace boundaries, sub-agent handoffs, tool wrapping, naming/tagging vocabulary) plus a Ratel-value/agent-health dashboard build-spec. Use to instrument Langfuse, wire up tracing, set up dashboards, decide where Langfuse goes or what to put on the board, onboard a just-signed partner, or `/ratel-langfuse-integrate`. Writes to <repo>/.ratel/; never edits code or calls the Langfuse API. The Langfuse branch of /ratel-observability-assessment; redirects if the vendor is LangSmith; routes to /ratel-langfuse-analyze once traces flow.
Read a customer's live LangSmith runs, traces, threads, and feedback scores via the LangSmith MCP server (or langsmith SDK fallback), pattern-match against the shared finding catalog, and write a findings report split into Ratel-flavored improvements and stack-agnostic low-hanging fruit. Use to scan the LangSmith dashboard, dig into the run tree or failed runs, review this week's runs, audit their observability, see what the agent's doing wrong, or `/ratel-langsmith-analyze`. Writes to <repo>/.ratel/ (accumulates); fails fast without a reachable server or API key, honest-skips clean data, never edits code. Follow-up to /ratel-langsmith-integrate.
Inspect an agent codebase and write a markdown plan wiring LangSmith tracing/observability (run-tree boundaries, sub-agent handoffs, tool wrapping, run-naming/metadata vocabulary) plus a Ratel-value/agent-health dashboard build-spec. Use to instrument LangSmith, add @traceable/wrap_openai coverage, group runs into threads, design dashboards or Monitor charts, decide where LangSmith goes, onboard a just-signed partner, or `/ratel-langsmith-integrate`. Writes to <repo>/.ratel/; never edits code or calls the LangSmith API. The LangSmith branch of /ratel-observability-assessment; redirects if the vendor is Langfuse; routes to /ratel-langsmith-analyze once traces flow.
Reconstruct your recent working context from your own chat history, live state, and the shared record (user reports, prior fixes, incidents), then hand back a tight current-state brief. Use for “recall my work on X”, “catch me up”, “what have I been working on”, “where did I leave off”, before starting or resuming work.
Register, add, or update existing custom Simulink block libraries and configure block policy and knowledge index. Use when a user wants to register, set up, configure, or add existing .slx library files for agent-assisted model building. Do NOT use when the user wants to create or author a new library from scratch — that requires building-simulink-models.
Configure pstack's provider-qualified models, per-family requested effort, and parent-owned routes per role. Verifies the assigned native and external lanes before writing the override sheet. Use for /setup-pstack, "configure pstack models", or changing pstack's model choices.
Configure an AI agent to send OpenTelemetry traces to Coval. Use when a user wants to add Coval tracing, instrument an agent for simulations or conversation monitoring, make traces show up in Coval, handle SIP/PSTN/WebSocket trace correlation, or replace the one-command wizard with a security-reviewable manual setup.
Set up or refresh UX Skills for a new or existing project by capturing product intent, learning users and evidence, the design system, engineering environment, accessibility expectations, terminology, and consequential decisions. Use only when the user explicitly asks to set up, initialize, connect, learn, or refresh UX Skills/project context; do not activate during ordinary UX work.
Keep a reviewable decision trail for long-running or unattended work: a TSV log with one row per decision (what, why, evidence, result). Local by default; commit it when a reviewer needs the trail to trust the result. Use for /show-me-your-work, autonomous or multi-phase runs, or work a human reviews after stepping away.
This skill should be used when the user asks to "organize the project rules", "clean up scattered documentation", "把规则迁移到 skills 目录", "优化 skill 路由", "提高 description 命中率", or "减少薄壳重复维护". Activate when a SKILL.md is too large, rules are duplicated across agent entry files, task routing or trigger_examples miss natural user language, or templates / thin shells / validation scripts need drift-resistant maintenance.
Use when the user needs to create a new Skill or update an existing Skill. This Skill provides complete workflow guidance for skill creation, including requirements analysis, writing standards, engineering-based construction, quality assessment, and iterative optimization.
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Write a well-formed agent SKILL.md from a workflow, SOP, or repeated procedure. Use when the user wants to "create a skill", "turn this process into a skill", "document this workflow as a skill", or asks why a skill isn't triggering.
Diagnose and improve an existing Claude/agent skill without changing its core purpose or original input/output contract. Use whenever the user wants to optimize, refine, clean up, debug, restructure, or improve an existing skill and needs contract-first review, hypothesis-driven proposals, representative test prompts, or conservative before/after validation instead of a blind rewrite.
Use when writing, condensing, or refactoring a BK-CI skill — for example tightening the `description`, unifying the main file skeleton, offloading heavy content into `reference/`, and settling progressive disclosure rules. Prefer this when the user wants to add or standardize a skill.
Manage skills, agents, extras, hooks, plugins, and MCP connection settings with the Skillshare CLI. Use when the user asks to configure or run Skillshare, install or sync resources across AI tools, manage shared memory notes, review a session for lessons or dream over saved memories, import MCP settings, manage targets, audit skills, recover backups, or troubleshoot Skillshare configuration and sync. Covers global and project modes, noninteractive automation, and guidance for the terminal UI.
Specification-driven implementation orchestrator — Coordinate spec-code, spec-review, and spec-test to implement from specs to PR. Reads issue-to-pr-workflow.md as playbook, delegates implementation to spec-code, review to spec-review, testing to spec-test. Manages fix loops, task progression, and PR creation. Does NOT write implementation code or perform reviews itself. English triggers: "Implement from spec", "Start implementation", "Execute spec tasks" 日本語トリガー: 「仕様書から実装」「実装を開始」「specタスクを実行」
Continue an approved Spec-backed workflow when the user says "implement", "go", "start", or "do it". After Codex Plan Mode, persist the next missing design.md or plan.json artifact and stop. When both artifacts exist, execute plan.json with TDD and report between batches.
Use when you need to execute I2 (implementation execution) in the sdlc-dev Spec Pack: implement in batches with `{FEATURE_DIR}/implementation/plan.md` as the single SSOT, run minimal verification, write back audit information, and report at batch checkpoints; stop immediately on blockers or open questions.
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.